DASHBOARDS & GRIDS
KPI vs metric
Every KPI is a metric. Almost no metric is a KPI. The difference is whether someone will act when it moves.
- Orderscount
- Revenuesum
- Average basketavg
- Returns rateratio
- Lines per orderavg
- Orders awaiting reviewcount
- Urgent orderscount
- Rejected orderscount
- Someone acts when it moves
- Compared with a period or a target
- Three to five on a dashboard, not thirty
Every KPI is a metric
A metric is anything you can measure about the business: orders, revenue, average basket, returns rate, lines per order, the number of orders awaiting review. There are hundreds, and a well-indexed data layer can produce any of them on demand. A key performance indicator is one of those metrics that has been promoted: somebody watches it, somebody owns it, and when it moves, something happens. Orders this period against last is a KPI for the sales director. Lines per order is a metric she might look at to explain the KPI. The number is not what makes the difference; the decision attached to it is.
The distinction is old - it predates dashboards by decades - and it survives because it is about attention rather than data. A business can measure thousands of things and watch five. The five are the KPIs. Everything else is context: useful when a KPI moves and you need to know why, noise when it sits in the top row pretending to be a signal.
The promotion test
Three questions turn a metric into a KPI, and a metric that fails any of them should stay a metric. Will someone act when it moves? If the answer is "we would want to know", that is a metric. If the answer is "the operations lead would call the warehouse", that is a KPI. Compared with what? 952 orders means nothing alone; 952 against the previous period, or against a target, is a signal. Who owns it? A KPI nobody is responsible for is a metric with a nice tile.
| Test | Metric | KPI |
|---|---|---|
| Can it be measured? | Yes | Yes |
| Will someone act when it moves? | Not necessarily | Yes, and they know who |
| Compared with what? | Often nothing | The previous period, or a target |
| How many on a dashboard? | As many as the widgets need | Three to five, at the top |
| Where it lives | In charts, grids and pivots | In a tile with its change |
| Example | Lines per order, returns rate, average basket | Orders this period against last; urgent orders right now |
How many a dashboard should carry
Three to five, at the top, in tiles. More than that and they stop being key: a row of fifteen tiles is a metrics list with large type, and the eye cannot tell which one changed. The operations dashboard in this site's examples carries orders in the period, urgent orders at 14 and rejected orders at 36, and that is enough to know whether the day is normal. The metrics behind them - by region, by category, by store - live in the charts and grids below, where they explain the KPIs rather than compete with them.
Comparison is what makes it a KPI
A KPI tile shows the number and its change against the previous period, and the change is the part people read. Up, down, by how much. Without it the tile is a fact; with it the tile is a prompt. That is why period comparison is a feature of the tile itself rather than a second chart: the comparison has to sit next to the number, where the glance lands. Choosing the right chart for business data covers how the comparison is drawn for the other chart types.
The comparison also needs to be the right one. Orders against last week is a signal for an operations desk; orders against the same week last year is the signal for a seasonal business; orders against target is the signal for a sales team. Same metric, three KPIs, depending on who is watching and what they would do. A tile should say which comparison it carries, because a number that is up against last week and down against last year supports two opposite decisions.
KPIs change, metrics do not
A metric is a definition and stays one. A KPI is a choice and gets revisited: during a stock problem, urgent orders is the KPI; during a quality problem, rejected orders is; in a growth quarter, new customers. The dashboard should make that promotion cheap - move a metric into a tile, give it a comparison - and make the demotion just as cheap, so the top row always carries what matters this quarter rather than what mattered when the dashboard was built.
A KPI also needs a cadence. Urgent orders is watched hourly; orders against target is watched weekly; customer retention is watched quarterly. Putting an hourly KPI and a quarterly one in the same row invites the wrong reaction to both. Group tiles by the rhythm at which someone will act on them, and the top row reads as a plan rather than a list.
The same distinction in chat
Ask an assistant for "orders this month" and the right shape is a KPI: one number, its change. Ask for "orders by region" and it is a leaderboard of metrics: North 395 · West 247 · East 168 · South 142. The assistant picks the shape from the question, which is the same judgement a dashboard designer makes once. How an answer gets its shape follows that choice.